mirror/OpenLLM
1
0
Fork 0
mirror of https://github.com/bentoml/OpenLLM.git synced 2026-06-12 10:29:36 -04:00
Code Issues Packages Projects Releases Wiki Activity

fix(yapf): align weird new lines break [generated] [skip ci] (#284)

Browse Source 
fix(yapf): align weird new lines break

Signed-off-by: aarnphm-ec2-dev <29749331+aarnphm@users.noreply.github.com>
This commit is contained in:
Aaron Pham
2023-09-01 05:34:22 -04:00
committed by GitHub
parent 3e45530abd
commit b7af7765d4
91 changed files with 811 additions and 1678 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
openllm-python/src/openllm/models/opt/modeling_flax_opt.py
View File

@@ -16,12 +16,11 @@ class FlaxOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Tok
__openllm_internal__ = True
def import_model(self, *args: t.Any, trust_remote_code: bool = False, **attrs: t.Any) -> bentoml.Model:
config, tokenizer = transformers.AutoConfig.from_pretrained(
self.model_id), transformers.AutoTokenizer.from_pretrained(self.model_id, **self.llm_parameters[-1])
config, tokenizer = transformers.AutoConfig.from_pretrained(self.model_id), transformers.AutoTokenizer.from_pretrained(
self.model_id, **self.llm_parameters[-1])
tokenizer.pad_token_id = config.pad_token_id
return bentoml.transformers.save_model(self.tag,
transformers.FlaxAutoModelForCausalLM.from_pretrained(
self.model_id, **attrs),
transformers.FlaxAutoModelForCausalLM.from_pretrained(self.model_id, **attrs),
custom_objects={'tokenizer': tokenizer},
labels=generate_labels(self))
@@ -45,6 +44,5 @@ class FlaxOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Tok
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors='np'),
do_sample=True,
generation_config=self.config.model_construct_env(
**attrs).to_generation_config()).sequences,
generation_config=self.config.model_construct_env(**attrs).to_generation_config()).sequences,
skip_special_tokens=True)

9
openllm-python/src/openllm/models/opt/modeling_opt.py
View File

@@ -18,8 +18,7 @@ class OPT(openllm.LLM['transformers.OPTForCausalLM', 'transformers.GPT2Tokenizer
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
import torch
with torch.inference_mode():
return self.tokenizer.batch_decode(
self.model.generate(**self.tokenizer(prompt, return_tensors='pt').to(self.device),
do_sample=True,
generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True)
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors='pt').to(self.device),
do_sample=True,
generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True)

16
openllm-python/src/openllm/models/opt/modeling_tf_opt.py
View File

@@ -11,18 +11,16 @@ class TFOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Token
def import_model(self, *args: t.Any, trust_remote_code: bool = False, **attrs: t.Any) -> bentoml.Model:
import transformers
config, tokenizer = transformers.AutoConfig.from_pretrained(
self.model_id), transformers.AutoTokenizer.from_pretrained(self.model_id, **self.llm_parameters[-1])
config, tokenizer = transformers.AutoConfig.from_pretrained(self.model_id), transformers.AutoTokenizer.from_pretrained(
self.model_id, **self.llm_parameters[-1])
tokenizer.pad_token_id = config.pad_token_id
return bentoml.transformers.save_model(self.tag,
transformers.TFOPTForCausalLM.from_pretrained(
self.model_id, trust_remote_code=trust_remote_code, **attrs),
transformers.TFOPTForCausalLM.from_pretrained(self.model_id, trust_remote_code=trust_remote_code, **attrs),
custom_objects={'tokenizer': tokenizer},
labels=generate_labels(self))
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
return self.tokenizer.batch_decode(
self.model.generate(**self.tokenizer(prompt, return_tensors='tf'),
do_sample=True,
generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True)
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors='tf'),
do_sample=True,
generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True)